department of aerospace science and technology · ‣part of the n/s arm of the northern cross...
TRANSCRIPT
Research activities
Space Missions Engineering laboratory
Teaching Staff: Franco Bernelli Zazzera, James Biggs, Camilla
Colombo, Pierluigi Di Lizia, Michèle Lavagna , Paolo Lunghi,
Mauro Massari, Francesco Topputo
Other: Amalia Ercoli Finzi (honorary professor)
• Public Insititutions
• ASI – Italian Space Agency
• ESA – European Space Agency
• CNES – French Space Agency
• DLR - German Space Agency
• Industries
• Thales-Alenia Space I
• Thales-Alenia Space F
• OHB System
• Kaiser-Traede
• OHB CGS
• QinetiQ
• Leonardo SES
• GMV Space Systems
• Elecnor Deimos Space
• SciSys
• CESI
• D-Orbit
• Vega-Telespazio
Collaborations
• Universities\Research Centres
• MIT – Massachusetts Institute of
Technology
• MSU – Michigan State University, US
• TAMU – Texas A&M University, US
• PU – Princeton University, US
• VirginiaTech, US
• NPS Monterey, US
• Purdue University, La Fayette, Indiana
• University of Arizona, US
• University of California, Irvine, US
• Southampton University, UK
• University of Surrey, UK
• University of Florida, US
• Tohoku University, Sendai, Japan
• Beihang, China
• National University of Defense
Technology, Changsha, China
• Shanghai Aerospace Control Engineering
Institute, China
• Universita di Milano Bicocca
• Universita Federico II-Napoli
• Università La Sapienza-Roma
• Politecnico di Torino
• Università di Bologna
• CIRA
• CNR
• INAF
Collaborations
ModellingSimulation
Analysis
&
Design
Global
Optimization
Robust\Optimal
Control
Muldisciplinary
optimization
Uncertainties
propagation
Soft Computing
Differential
Algebra
Space
Environment
Multi-body
Dynamics
Mission Analysis
Space Science & Technology
Relative
Dynamics
Local
Optimization
Dynamics
&
Control Adaptive
Control
Atmospheric
Maneuvering
Mission Analysis
and
Interplanetary
trajectories
Planetary Orbits
Precise Station
Keeping, SAA, contact
windows, etc
Mission Analysis
Robust orbits
propagation
Entry Descent
&LandingFormation Flying
Control
Irregular natural
bodies geopotential
modelling
N-Body models
trajectories
Orbit determination
under uncertainties
Space Situational Awareness – NEO and Debris
Algorithms implementation
• Orbit determination based on
different sensors/sensors architecture
• Accurate propagation of uncertainties
• Conjunctions identification
• Collision probability computation
System trade-off and design
• Definition and critical review of
requirements for sensors/sensors
architecture implementation
• Support in the design of sensors
and network architecture (e.g. sensor
type, sensor performances,....)
• System simulation and assessment of system performances
Formation Flying
Absolute and relative trajectories design
Station-keeping for both impulsive and low-thrust maneuvers
Reconfiguration maneuvers with optimal feedback control
4 s/c reconfiguration 4 s/c reconfiguration (collision
avoidance)
Atmospheric Phases
Multidisciplinary and Multiobjective optimizationMultiobjective optimization of AGA
Multidisciplinary optimization of EDL
phase
Aero Gravity Assist and Aerocapture maneuvers
trajectory, control, and vehicle shape optimization
EDL phases:
guidance, shape, thermal
protection, sequence timing
Orbit Design - Interplanetary
Global optimization
– Deterministic optimization of
MGA+DSM transfers (gravity assist
space pruning based on Differential
Algebra)
Objective function (Earth-Mars
transfer)
Low-thrust transfer (EMMJ)
Local optimization
– Optimization of MGA+Low-Thrust
Goal: Validate GNC landing, vision based, for planetary and low gravity bodiesdescent
Features: precise landing; retargeting landing (adaptive guidance) site upon hazard infos,on board closed loop
Automated Guidance, Navigation and Control
for Spacecraft Landing
In red: building blocks under development @ Politecnico di Milano DAER
Adaptive Guidance
Landing trajectory computation: Constrained Two Points Boundary Value Problem.
Requirements:
• Fast computation (real-time trajectory update);
• Fuel optimal (in case more than one retargeting is required during the descent).
Semi-analytical, optimized landing trajectory:
• Polynomial formulation satisfies boundary constraints;
• Additional parameters can be tuned to optimize the landing trajectory, according to
path constraints;
• Light direct optimization algorithms (eg. Compass Search);
• Tested through Monte Carlo simulation for both fast (planetary) and slow (asteroids)
dynamics.
On going activity
Closed-loop simulation campaign to assess the capability of hazard avoidance, navigation
and guidance to work together)
Automated Guidance, Navigation and Control
for Spacecraft Landing
Automated Guidance, Navigation and Control
for Spacecraft Landing
Moon landing adaptive guidance MC test case
• Diversion ordered at altitude 2000m, 600m (1σ);• dispersion at touchdown <16m (3σ). Navigation errors included
Asteroid landing adaptive guidance MC test case
Mission Analysis - Earth
Mission design for EO missions including
– Orbit selection, launcher selection, launch windows
– Coverage, Access, and Light condition analysis
– Orbital simulation and environmental analysis
– ∆V Budget and station keeping strategy
– Optimal control theory applied to station keeping
SZA vs day of the year Projection of Earth orbits
Dynamics and Control of Non-Keplerian Orbits
16
GoalLow Energy Interplanetary Transfers (LEIT)
Solution
The R3BP to model the space dynamics
Transit orbits of the Sun-planet systems
Low energy planet approach through small necks
Invariant manifolds as key to design LEIT
4BP split in two R3BP
Intersection of the invariant maifolds in the solution
space
The restricted three-body approximation as
refinement of the patched-conics method for LEIT
design
SMALL BODIES GRAVITY FIELD MODELS
Accurate mass distribution/gravity field
model of the body
Effective trajectory design around a low gravity bodyGravitational potential
Acceleration model
SMALL BODIES GRAVITY FIELD MODELS
Asteroid gravity field models
GRAVITATIONAL AGGREGATE
Used to simulate aggregation process. Very accurate for
certain classes of asteroids. It models internal voids
ELLIPSOIDSimple model, it works well to
design trajectories far from asteroid surface
POLYHEDRONVery accurate model, it works
well in the whole domain, including the design of
landing/lift off trajectories.Customizable on the specific
asteroid
SMALL BODIES GRAVITY FIELD MODELS
Example: Ellipsoid vs Polyhedron
latitu
de
[d
eg
]la
titu
de
[d
eg
]
longitude [deg]
longitude [deg]
a [m/s2]
a [m/s2]
a [m/s2]
a [m/s2]
[km]
[km]
Gravity field is studied and compared between different
modelsTo be validated through data on subsurface
composition , material distribution and structure
SMALL BODIES GRAVITY FIELD MODELS
Example: N-body aggregation sequence
Gravitational aggregate
1 2 3
4 5 6
Asteroids and binary systems formation\distruption and rotational dynamics understanding
To be validated through data on subsurface
composition , material distribution and structure
Space Debris Problem
‣ Debris: any defunct object in Earth orbit
• Size ranging from mm to a few meters
• More than 22,000 objects larger than 10
cm currently tracked
• Explosions and collisions boost the
number of fragments and could trigger
Kessler’s syndrome
Explosion
Collision
‣ Crucial: Monitoring, tracking, and
predicting trajectories of space debris
• Identification of orbital collision for
avoidance maneuver planning
• Re-entry forecast for larger objects
• Detection of in-orbit fragmentations
and collisions between uncontrolled
objects
‣ Highest concentration: LEO and GEO Credits: ESA
Orbit Determination with Multibeam Radars
E/W arm: single antenna
(564 m x 35 m)
N/S arm: array of 64 antennas
(640 m x 23.5 m)
‣ Part of the N/S arm of the Northern Cross Radiotelescope (Medicina,
Bologna, Italy) has been refurbished to enable radar multibeaming
• 32 receivers mounted on 8 antennas
• Resulting FoV = 38 deg2 (Dec: 5.7 deg, AR: 6.6 deg)
• Back end processing allows the FoV to be divided into 32 beams
Main advantage: beams illumination sequence supplies declination and right
ascension profiles during transit in addition to range and Doppler shift
Orbit Determination with Multibeam Radars
‣ Part of the N/S arm of the Northern Cross Radiotelescope (Medicina,
Bologna, Italy) has been refurbished to enable radar multibeaming
• 32 receivers mounted on 8 antennas
• Resulting FoV = 38 deg2 (Dec: 5.7 deg, AR: 6.6 deg)
• Back end processing allows the FoV to be divided into 32 beams
Main advantage: beams illumination sequence supplies declination and right
ascension profiles during transit in addition to range and Doppler shift
BEST-2
Orbit Determination with Multibeam Radars
BES
T-2
Step 1 Step 2
‣ Two-step algorithm developed for orbit determination:
• Step 1: Estimation of right ascension and declination profiles from SNR profiles
• Step 2: Estimation of object position and velocity by matching orbital trajectory
with range measurements and observables profiles
‣ First numerical tests suggests that orbit determination could be performed
with a single transit in the FoV (good RCS estimate necessary)
‣ The software has been extended to manage both radar and optical inputs
‣ Software currently under developments for INAF under ASI contract
Correlating Observations
‣ One optical/radar observation of an unknown object is not sufficient to
accomplish an initial orbit determination
‣ However, an admissible region (AR) can be identified from one
observation based on energetic constraints
‣ The AR must be propagated forward to schedule additional observations of
the same object
Intensive Monte Carlo
simulations
‣ Moreover, assume two ARs are
available from two different
observations
• How can we check if the two
observations are correlated?
AR 1AR 2
T11T21
Same object?
1
• Main drawback: the region is usually
large and the dynamics is nonlinear
Correlating Observations
AR 1AR 2
T1 T21
AR
2
Tcommon1
‣ Main Idea: use differential algebra and automatic domain splitting to
propagate large admissible regions
• Differential algebra is used to replace pointwise numerical integrations with the
fast evaluation of polynomials
• Automatic domain splitting is added to improve accuracy of polynomials for very
large ARs and/or long term propagations
‣ Main advantages
• The ARs can be quickly propagated to schedule new observations
• ARs from different observations can be propagated to a common epoch and
their intersection can be checked to correlate the observations
‣ Under study for Air Force R.L. in partnership with Universidad de La Rioja
Ballistic captureFor Mars missions
27
−8 −6 −4 −2 0 2 4
x 10−3
−7
−6
−5
−4
−3
−2
−1
0
1
2
3
x 10−3
x (adim.)
y (
adim
.)
Ballistic capture transfers to Mars:
•Can achieve substantial savings in capture Δv (~20%)
•Is safer than hyperbolic approach, allows multiple
insertion options
•Involve much more flexible launch windows
Space SHIPSpace Systems with Hybrid Propulsion
28
Attempt to combine the benefits of
• Chemical propulsion
• Electric propulsion
Can be applied in missions to
• Moon
• Mars
• NEOs
Preliminary solutionCPU model
Thruster model Refined solution
• Overcome the binary trade space
• Fully-chemical and fully-electrical
configurations viewed as special hybrid
solutions
• Challenge is to assess their implication
at system level
Space SHIPSpace Systems with Hybrid Propulsion
29
• Propulsion not expressive innovative, their combined
utilization brings novelty
• Hybrid system can ease operation (no GA)
• High power needed during the transfer, but
• Useful for high-power payloads
• Which subsystem do the SA belong to?
Preliminary results show:
• Considerable savings wrt fully chemical solutions
• Shorter transfer times wrt fully electric solutions
• Possible standardization for Moon, Mars, NEOs mission
FSM: Final spacecraft mass
UMAT: Useful mass at target(Results from ITT 6791)
Miniaturization
Guidance and control of resource limited
spacecraft
Developing guidance and contol algorithms for highly constrained but highly
responsive spacecraft
Computationally efficient for on-board implementation.
Guidance methods that incorporate 6 dof obstacle avoidance and precision docking.
Control methods for the deployment in uncertain environments.
Fault-tolerant control methods.
Algorithms for efficient motions using novel propulsion – solar sail, pulse-plasma thrusters
More responsive, cheaper.
Less reliable, resource limited.
Tools
Nonlinear Dynamical systems and Control theory.
Differential geometry for global motion planning on non-Euclidean spaces.
Simulation.
Test-bed experimentation.
Dynamics in highly inhomogeneous gravity fields
Investigations of the dynamical environments about Asteroids and the moons of Mars
Analytical Perturbative methods for global orbit identification.
Numerical continuation to compute families of orbits.
Nonlinear and linear stability analysis of orbits.
Station-keeping in the presence of uncertainties and constraints.
Modelling and simulation of spacecraft
dynamics
MODELICA: a language for physical systems modelling.
Main features:
– Based on equations rather than assignments;– Acausal models;– Can be used to model multidomain systems;– Object-oriented language (based on classes);– Component interaction described by connectors;
Modelica Spacecraft Dynamics Library:
– Based on the DLR Modelica Multibody Library;– A unique environment for the entire AOCS design cycle;– Possibility of multidomain simulation;– Easy management of spacecraft configurations; – Rapid prototyping of AOCS subsystems.
Generic spacecraft simulator (Modelica/Dymola)
Sensors block
Spacecraft model
Spacecraft dynamics
Actuators block
Data sheets
Attitude determination:
estimation and filtering
Background:
– development of the EKFs for the Italian missions MITA and AGILE.
– Development of a novel algorithm for
star identification; implementation of
on-board code for star sensor-based
attitude determination.
Recent activities
– Study of UKFs for attitude estimation
– Predictive filters for attitude estimation
– Globally convergent filters for magnetic attitude estimation
– High accuracy attitude and rate estimation using magneto-hydrodynamic sensors
Star cameraStar
Identification
Attitude
Determination
Star Catalog
Star sensor
Attitude
Angular
rate
Starsensor
0 2000 4000 6000 8000 10000 12000-0.04
-0.03
-0.02
-0.01
0
0.01
0.02theta
time [s]
Roll
angle
[ra
d]
0 2000 4000 6000 8000 10000 12000-0.02
-0.01
0
0.01
0.02psi
time [s]
Pitch a
ngle
[ra
d}
Attitude control:
full and partial magnetic actuation
Problem:
attitude control for satellites using magnetic coils as primary or sole actuators
Torque generation mechanism: interaction between current-driven coils and geomagnetic field. Magnetic field periodic along orbit LTP linearised attitude dynamics.
Critical issues:
Magnetic actuators are intrinsically time-varyingThe magnetically actuated spacecraft is not instantaneously controllable
Magnetic attitude control
Nominal phase: tools for numerical optimisation of linear state feedback controllers
Acquisition phase: methods for convergence analysis of acquisition transients
Partially magnetic control: design tools for coils+ thrusters configuration
Robust analysis and synthesis: methods and tools for robust design and worst-case analysis for linear magnetic ACS loops.
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50.9997
0.9998
0.9999
1
q1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.01
0
0.01
0.02
q2
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.02
0
0.02
q3
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.01
0
0.01
q4
Orbits
Space ShepherdMonitoring refugees in the Mediterranean Sea
38
•Loss of human lives
•Emergency conditions
•Raising lands
•Critical shelter
Motivations
• Land systems used
• Local coverage
• Deployment of assets
• Territorial waters
State of the art
To use satellites already operating on a regular basis to monitor, detect, track
immigration flows in the Mediterranean Sea and support rescue and landing
The idea
• Innovative approach
• Global coverage
• Dual use (“zero” cost)
•Multidisciplinary nature
Originality
Context
2014: 170k migrants (460/day)
Jan-Apr 2014/Jan-Apr 2015:
+2900% casualties
2015: 240k migrants (estimated)
Space ShepherdMonitoring refugees in the Mediterranean Sea
39
•To remotely monitor portions of sea
•To detect migrant vessels
•To infer the situational awareness
•To track the vessels and raise flags
•To support the rescue operations
•To improve logistics of landings
•To develop a simulation framework
•To integrate SAR/optical imageries
GoalsMethodology
Land radar range 100
km (over-estimated)
Sea radar range 60
km (over-estimated)
Ship-to-ship visual range
20 km (over-estimated)
Space ShepherdMonitoring refugees in the Mediterranean Sea
40
•To remotely monitor portions of sea
•To detect migrant vessels
•To infer the situational awareness
•To track the vessels and raise flags
•To support the rescue operations
•To improve logistics of landings
•To develop a simulation framework
•To integrate SAR/optical imageries
GoalsMethodology
Land radar range 100
km (over-estimated)
Sea radar range 60
km (over-estimated)
Ship-to-ship visual range
20 km (over-estimated)
Space ShepherdMonitoring refugees in the Mediterranean Sea
41
•To remotely monitor portions of sea
•To detect migrant vessels
•To infer the situational awareness
•To track the vessels and raise flags
•To support the rescue operations
•To improve logistics of landings
•To develop a simulation framework
•To integrate SAR/optical imageries
GoalsMethodology
Land radar range 100
km (over-estimated)
Sea radar range 60
km (over-estimated)
Ship-to-ship visual range
20 km (over-estimated)
4 COSMO-SkyMed,
1 day, resolution 30 m,
swath 100 km
Space ShepherdMonitoring refugees in the Mediterranean Sea
42
•To remotely monitor portions of sea
•To detect migrant vessels
•To infer the situational awareness
•To track the vessels and raise flags
•To support the rescue operations
•To improve logistics of landings
•To develop a simulation framework
•To integrate SAR/optical imageries
GoalsMethodology
Land radar range 100
km (over-estimated)
Sea radar range 60
km (over-estimated)
Ship-to-ship visual range
20 km (over-estimated)
Space ShepherdMonitoring refugees in the Mediterranean Sea
43
TLE
SGP4
Sensor geometry
Ground stations
Contact windows
Kinematics model
Initial uncertainty
Uncertainty propagation
Longitude10 15 20 25 30 35
La
titu
de
30
32
34
36
38
40
42
C28
C39
??-5
C56
C62
C63
C7
C17
C29
C42
C51
C58
C61
C13
C26C17
C21
C44
C46
C56C40
C61
??-8
COSMO-SkyMed 2
COSMO-SkyMed 3
COSMO-SkyMed 4
COSMO-SkyMed 1
SAR strips @ sea
Optical images @ ports
•Monitoring
•Tracking
•Support to S&R
• Warnings
• Info on targets
• Info on nearby vessels
• Feedback
Type of ship Length [m] Number of people carried
(from sources)
Speed range
[knots]
Source
Rust-bucket
Old african fishing boats
15-20 100-200 2-10* www.marina.difesa.it
www.lastampa.it
Fishing motorboats 25-30 200-250 (even 350)
2-10 * www.ilsecoloxix.it www.ilfattoquotidiano.it
Rubber dinghies 10-15 100 5-10* www.marina.difesa.it
Big fishing
motorboats or small cargo boats (Mother
ships)
≥ 30 >200 (even
500)
10-15
(presumed)
www.gdf.gov.it
www.ilsole24ore.com
Cargo or merchant ships (for the whole
crossing)
30-70 ≥ 500 200-800
10-20 * www.palermo.repubblica.it www.repubblica.it
Sail boats 10-20 Variable ( > 40) < 10
(presumed)
www.gdf.gov.it
www.europaquotidiano.it
www.corriere.it
Analisi del sistema AIS Capire il funzionamento del sistema AIS (Automatic Identification System) è di fondamentale importanza ai
fini del progetto in quanto le immagini satellitari devono essere depurate delle imbarcazioni autorizzate, che trasmettono la propria posizione attraverso il sistema AIS. Per questo motivo è stata condotta un’analisi sul
funzionamento di tale sistema. I risultati sono descritti nel report 1 “project meeting #1” (allegato) e sono brevemente riassunti di seguito:
· In Italia il sistema è obbligatorio sulla quasi totalità delle imbarcazioni. Nello specifico, le imbarcazioni autorizzate che presumibilmente saranno presenti nelle immagini satellitari sarano tutte
dotate di sistema AIS;
· Sebbene esista una sperimentazione sull’uso dei satelliti, il sistema AIS non utilizza una
trasmissione satellitare, bensì una tecnologia peer-to-peer, per cui è possibile trasmettere la propria posizione solo se si è in vista di altre imbarcazioni che possano fungere da “ponte” per il segnale.
AIS-Problem1: Missing Ships
19/03/15 Space Shepherd
Esempio di posizionamento con sistema AIS
Requisiti, architettura e implementazione del simulatore Lo sviluppo del simulatore è stato messo in atto attraverso i seguenti passi: 1) Creazione di un database di
satelliti candidati (allegato), 2) formulazione dei requisiti, 3) definizione dell’architettura, 4) validazione
preliminare. Per quanto riguarda i requisiti, il simulatore opererà in tre diverse modalità, a seconda del
Space ShepherdMonitoring refugees in the Mediterranean Sea
44
Coverage Mean Time
Mean Cycle Time
Average Cycle Response Time
Mean time between two
acquisitions of the same
target
Mean to:1) uplink the
instructions, 2) take the
image, 3) download the
image to the G/S
Mean time to: 1) plan the
acquisitions, 2) uplink
the instructions, 3) take
the image, 4) download
the image to the G/S, 5)
post-process the data
Satellite groups
Estimated cost
Features
• Limited costs, already-existing assets exploited
• Increased situational awareness “for free”
• System flexible and scalable
• Scientific and commercial satellites only
• Integrates (cannot replace) in-situ S&R operations
• Agreements among operators are necessary
Navigation
&
Environment
Guidance
&
Control
Facilities
&
Experiments
Space Systems
Space Science & Technology
Space Systems
Digital
Environment
reconstruction
Visual Odometry
for rover
Navigation
Hazard Maps
generation for landing
Planetary soil
sampling toolsSurface
Locomoton
Planetary ISRU
plants
Space platform
and orbital
robotics design
Space Science & Technology
Spinning Target 8t
(Envisat)
Launch and capture
In orbit Robotics: Analysis & DesignActive Debris Removal
•Dynamics of complex flexible system in space
•Impact and wrapping, pulling
NUMERICAL SIMULATORS:
Design & Dynamics
Contact devices
Flexible systems=net+tether
In orbit Robotics: Analysis & DesignActive Debris Removal
•Stack flexible dynamics control during disposal
NUMERICAL SIMULATORS:
Design & Dynamics
Contact devices
Flexible systems=net+tether
Free dynamics + RCS
stabilisation
Large powered
maneuver
In orbit Robotics: Analysis & DesignActive Debris Removal Parabolic Flight
BREADBOARDING and ExperimentsSuccessful zero-g experiment: Novespace Flight in Bordeaux, June 12, 2015
Surface Hazard Maps autonomous generation to :
safe landing on unknown objects\surfaces
multi-level clustering algorithm supported by supervised learning (1024x124 tcom=3s)
In Orbit Robotics: Navigation &
Environment\Guidance Autonomous GNC for landing
Self-organizing
maps
Multilayer
Neural Network
Visual-based Navigation
•Single camera relative navigation.
•Image features are extracted and tracked
between subsequent frames.
•Data fusion together with complementary
sensors: Inertial Measurement Units
(IMU) and laser altimeters.
•On going.
In Orbit Robotics: Navigation & Environment\Guidance Autonomous GNC for landing
Comet 67P landing site-Hazard map
Moon – Lamror Q Crater
Backup landing region from NavCam –
Rosetta s\c
In Orbit Robotics: GNC for landing Facility setup
Moon – Facility architecture Moon DEM for the diorama and
rendering
Manufactured Diorama Visual results
Biomimetic Legged robot projects inspired by insects:
Prototype to
CTRNN Neural Network and distributed control
testing
Smart Materials for actuation
Surface Robotics: locomotion
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50
5
10
15
20
25
30
35
40
45
T [Nm]
DP
[N
]
Drawbar pull as a function of torque with different eccentricity
e = 0
e = 0.8
Wheeled Rover Traversability
improvements
Sensitivity to wheel shape variation based
on Smart materials implementation
-0.1 -0.05 0 0.05 0.1-0.1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0.1
x [m]
z [m
]
Deformation from circular to elliptic wheel
increasing e
– Digital Environment Reconstruction: Stereo Vision System
– Motion State Reconstruction: Visual Odometry
– Path planning algorithms
z
mx
m
y
Displacement
And Rotation
55
Navigation & Environment: Surface mobility
Robotics: ControlHigh level control: autonomous reasoning
56
Deliberate
GoalActivity
Execute
Actioncommand
React
MonitoringBehaviour
Three layers architecture
Goal Manager
Recovey
Agent
Planner/Scheduler
Executer
Failure
Detection
Isolation
WorldActuatorsSensors
World
Environ.
GM
P/S
EX
Agent 1
FDI
R
HW1 M
W
Simulator
M
COM COM
HWn
GM
P/S
EX
M
Agent n
FDI
R
Logic Single Agent Multi-Agent
Applied to:
•Physical distribution scenarios: team of homogeneous\heterogeneous space vehicles
•Functional distribution scenarios: ground stations functionalities for data management
Surface Robotics: Analysis & DesignSampling mechanisms
SD2 Soil modeling DEM
Sampling tool design
Granular mechanics: simulation of sampling
Instant pusher
• Fast grab bucket
1,5s; 37g
Rotating stinger
• 1,5s; 7g
1,5s; 58g
3s; 38g
2014 Philae – Rosetta Lander – landed on comet
Churymov-Gerasimenko: PoliMi drill SD2 PISD2
Surface Robotics: Cometary lander - Philae
Electric power generation
simulator
Mechanical verification
Drill behavior in uncertain material and
microgravity conditions
Definition of optimal drilling strategies
Definition of contingency operations
Scientific use of SD2
Mission plans development
2014 Philae – Rosetta Lander – landed on comet
Churymov-Gerasimenko: PoliMi drill SD2 PI
SD2
Surface Robotics: Planetary sample collection
Current evolution
Drill for Moon caps: behavior in icy soil
Definition of optimal drilling strategies to
preserve volatiles
Energy exchange modelling and
validation
Subsurface composition and
physical parameters needed to
tune the tool and operations
design
Surface Robotics: Biomimetic Legged Robot
Until Now Planetary Exploration has been carried on with wheeled rover
Wheels have problems:Overcoming of large obstacles
Navigating on high slope
Low efficiency on rough terrain
Legs can solve those problem but introduce other:Mechanical configuration not passively stable
Higher power consumption (even at rest)
Low PL/BUS Mass Ratio
The Walking Motion of insects is usually characterized by:
Non-continuous contact with ground
Multiple DOFs should be coordinated and moved together
Intrinsic redundancy (required by passive stability)
Walking Motion Generation and Control is realized with a Decentralized approach (Reflex-based): The periodic motion is not imposed centrally but arises from the basic behaviors (reflexes) implemented at leg and joint level.
Legs CoordinationLegs Control
Joint Control
Decentralized: Bottom-up Approach
Centralized
Surface Robotics: Biomimetic Legged Robot
The controller is organized in two layers:
Low Level Velocity Joint control (RT embedded on robot HW)
High Level Motion Generation and control
– Single Leg Control (Dynamic CTRNN)
• Reflex included
– Leg Coordination (Fuzzy Logic Rules)
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
x [m]
z [m
]
Deformation from circular to circular wheel
increasing r
Surface Robotics: Variable shape Wheel
Solutions in planetary rovers to improve traversability performances:
Grousers
Flexible wheels
Variable shape wheels
Eccentricity variation (e) Radius variation (r)
-0.1 -0.05 0 0.05 0.1-0.1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0.1
x [m]
z [m
]
Deformation from circular to elliptic wheel
increasing e
Traversability is the robot’s ability to traverse soft soils or hard ground without loss of traction
Surface Robotics: Variable shape Wheel
Index of traversability is drawbar pull
DP = H - R
DP = 0
NO MOTION
DP > 0
MOTION
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.50
5
10
15
20
25
30
35
40
45
= 0°
= 5°
= 10°
= 15°
= 20°
T [Nm]
DP
[N
]
Drawbar pull as a function of torque with different slope angle
0 0.5 1 1.5 2 2.5 3 3.5 40
5
10
15
20
25
30
35
40
= 0°
= 5°
= 10°
= 15°
= 20°
T [Nm]
DP
[N
]
Drawbar pull as a function of torque with different slope angle
Circular wheel (e = 0) Elliptic wheel (e = 0.8)
≈14° ≈21°
Surface Robotics: Variable shape Wheel
0 1 2 3 4 5 6 7 8 9 10 110
5
10
15
20
25
30
35
40
45
50
55
r = 0.06m
r = 0.09m
r = 0.12m r = 0.15m
r = 0.18m
T [Nm]
DP
[N
]
Drawbar pull as a function of torque with different radius
Sampling Mechanisms: DEM
‣ Soil is modelled as a granular material: set of particles interacting with
contact or contactless forces
‣ At each integration step:
1. Set forces on particles from previous step
2. Detect collisions and update interactions
3. Solve interactions to apply forces
4. Integrate motion equations to study evolution
‣ Main advantages:
• Reduced cost for preliminary design: e.g., performance comparison between
different mechanisms to limit number of breadboards
• Provides requirements for system design (mechanical loads)
• Can be used to simulate behaviour in conditions that are costly or difficult to
reproduce in test campaigns (e.g., microgravity)
‣ The model is first calibrated to represent the
target soil
Sampling Mechanisms: DEM
‣ Main phases:
• Soil specimen generation and calibration:
o Set soil properties (density, friction angle,
Young's modulus,…)
o Soil deposition, shaking, and scraping
• Sampling tool modelling
o Geometry
o Mechanical properties
• Kinematic trajectory generation
o Kinematic trajectory can be imposed to
reduce computational cost
o Presence of springs can be simulated with
triggers
o Tools dynamics can be included at
increased computational cost
• Simulation
‣ Study performed for Selex-ES under ESA Contract “Breadboard of a
Sampling Tool Mechanism for Low Gravity Bodies” (E915-003MS)
Sampling Mechanisms: DEM
‣ Study performed for Selex-ES under ESA Contract “Breadboard of a
Sampling Tool Mechanism for Low Gravity Bodies” (E915-003MS)
spring
‣ Main phases:
• Soil specimen generation and calibration:
o Set soil properties (density, friction angle,
Young's modulus,…)
o Soil deposition, shaking, and scraping
• Sampling tool modelling
o Geometry
o Mechanical properties
• Kinematic trajectory generation
o Kinematic trajectory can be imposed to
reduce computational cost
o Presence of springs can be simulated with
triggers
o Tools dynamics can be included at
increased computational cost
• Simulation
Sampling Mechanisms: DEM
‣ Preliminary results
1-g simulation 0-g simulation
Simulation Filling percentage Collected sample [g]
1-g 20.0 % 33.0
0-g 29.2 % 48.5
Sampling Mechanisms: DEM
‣ Preliminary results
1-g simulation 0-g simulation
Simulation Filling percentage Collected sample [g]
1-g 20.0 % 33.0
0-g 29.2 % 48.5
Sampling Mechanisms: DEM
‣ Preliminary results
1-g simulation 0-g simulation
Simulation Filling percentage Collected sample [g]
1-g 20.0 % 33.0
0-g 29.2 % 48.5
Sampling Mechanisms: DEM
‣ Preliminary results
1-g simulation 0-g simulation
‣ DEM could serve other fields. E.g., simulation of landing on dusty soils:
• Estimate mechanical loads in different landing configurations and for different
landing mechanisms
• Study the dynamics of dust particles to estimate the effects on mechanisms or
the risk of contamination
Experimental Facilities
• Thermal Vacuum Chamber (-75° +200°, 10-6 mbar)
• Palamede:students’microsatellite implementation
• Friction free table
• Exploited for Internal and
external testing
• Predictive control to optimize
chamber performance
• 42,8x42,8x40 cm Earth
observation microsatellite
• p\l: CCD camera; Ptot=58W;
Mtot=35kg
• Educational and research goals
• 3x3m glass table
• GNC for proximity maneuvering
• robotics in microgravity conditions
Generalized Predictive Control: TVC tests
Model Predictive Control•Mathematical model of the System (predictor)•u(t) obtained with minimization of a cost function
•Recursive strategy (update of prediction at each step)
Identification (ARX Method)•Linear multistep predictor•Identification can be performed online or offline
s s p py k Tu k Bu k p Ay k p
k tk p sk h
:
:
sh
p
Prediction horizon
Order of the model
1
2
T T
s s s s s sJ k y k y k Q y k y k u k Ru k
1 k t
,u y
u
y
passato futuro
k p sk h
Predicted output
evaluation( )u k
Desired output
Generalized Predictive Control: TVC tests
Performances WRT classical control
TRP on the component under test
Preliminary tests
– constant error (can be corrected)
– Regular control (lower value)
PID
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
x 104
58
59
60
61
T [
°C]
t [s]
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
x 104
0
2
4
6
8 VPID
[V
]
TTRPo
TTRP
VPID
GPC (SIMO)
TRP
TBASEPLAT
E
y
TRP
TBASEPLAT
E
1y2y
4000 5000 6000 7000 8000 9000 10000 11000 12000 13000 14000
58
59
60
61
T [
°C]
t [s]
4000 5000 6000 7000 8000 9000 10000 11000 12000 13000 140000
2
4
6
8
VR [
V]
TTRPo
TTRP
VR
8
5s
p
h
0.005
1dt s
Generalized Predictive Control: TVC tests
Variation of TRP during test
New identification maintain performances
Without identification the controls fails.
Ph.D. on the application of Nonlinear GPC to attitude control of unkownobject (Uncooperative Targets, Debris Removal)
TRP2TRP1
( *)y per t t( *)y per t t
500 1000 1500 2000 2500 3000 3500 4000
37.5
38
38.5
39
39.5
40
40.5
41
T [
°C]
t [s]
500 1000 1500 2000 2500 3000 3500 40000
2
4
6
8
VR [
V]
ID
t = t*
TBPo
TBP
VR
TRP1 TRP2
New–Idendification
1000 1500 2000 2500 3000 3500 4000 4500 500036
37
38
39
40
41
T [
°C]
t [s]
1000 1500 2000 2500 3000 3500 4000 4500 50000
2
4
6
8
VR [
V]
t = t*
TTRPo
TTRP1
TTRP2
VR
NO
Identification
TRP1 TRP2
Feasibility studies
3 per year phase A studies on potential future space mission of ESA interest
•Europa Lander; Binary Asteroids deviation; Formation Flying at the magnetopause
•Manned mission to NEA; Fractioned satellite on Earth orbit; Phobos sample return
•Mission to Enceladus; Mission to Pluto-Caron; Active Debris Removal & in orbit servicing
•Lunar cold trap mission; Mission to Neptune; Mars Lagrangian point station
•Venus sample return; Pioneer Anomaly detection mission; Troposhere monitoring
•Jupiter Moons tour; large X-ray telescope; Earth Energy sources mapping
• Manned mission on the Moon; GPS on Mars; Vega Launcher upper stage
•In situ Resource Utilisation on Mars; interplanetary GPS;